Grid-Based Problem Solving Environments:
Implications for Development and Deployment of Numerical Software

In 1984 W. J. Cody reviewed progress in numerical software during the previous two decades and then identified future challenges posed by the rapid advances in computing technology*:

1. decline of the local central computing facility
• isolated from research
• libraries neglected
2. rise of personal computing
• isolated from everyone
• poor software
3. rise of remote computing facilities
   • good libraries
   • resident specialists

Cody’s foresight over 20 years ago was remarkable. The establishment of supercomputer centers coupled with development of distributed computing, including the recent development of grid infrastructure, has resulted in users being increasingly dependent on software resources on remote systems.

The IFIP Working Group (2.5) on Numerical Software is organizing an IFIP Working Conference on Grid-Based Problem Solving Environments: Implications for Development and Deployment of Numerical Software. In this context, the phrase “Grid-Based Problem Solving Environments” is synonymous with “science gateways” or “science portals”, nomenclature introduced recently by the grid community. The conference will bring together members of four communities:

• users of both grid-based and traditional problem solving environments;
• developers of both grid-based and traditional problem solving environments;
• developers of grid infrastructure; and
• developers of numerical software

for a week of intensive interaction, July 17-21, 2006, in Prescott, AZ. The use of the Internet to bring together providers and users of resources has become commonplace. In fact, one of the earliest such tools was Netlib, which started in the 1980’s as an email-based application for distributing numerical software. Other services, such as NetSolve (for linear algebra) and NEOS (for nonlinear and optimization problems) have demonstrated the potential of grid-based problem solving environments from the numerical software perspective. Domain specific PSEs, such as Cactus (developed for numerical relativity) and PYRE (developed for shock physics), have been applied outside their original discipline to generate new PSEs and are being extended to exploit grid technology. Likewise, many tools exist for making a service available. However, there is more to making use of a grid-based service than simply knowing its web address. Does the service work reliably? What are its limitations? Can it be combined with other services? There are also opportunities for improving the ability of applications to use the best numerical software, for example, by simplifying the acquisition and use of high-quality numerical software. The development of numerical software can benefit from the experience of the scientific and engineering communities using and developing new grid-based PSEs, for example, defining interfaces more appropriate for integrating numerical software into grid-oriented applications and exploiting test sets and tools for comparing different methods.

Issues of interest to this working conference include, but are not limited to:

• accuracy contracts and software services;
• standards for problem specification;
• service models for the use of numerical software;
• using the grid to link numerical and other services together;
• experiences with web-based numerical services;
• application-oriented numerical interfaces such as web portals;
• software deployment issues including updates and bug fixes;
• large data (including data security) and grid-based numerical software;
• grid-based services as an alternative to deployment; and
• evaluation and comparison of both production and research software.

This conference will build upon the experience and insights gained during past working conferences organized by WG2.5, in particular, “WoCo4: Problem Solving Environments for Scientific Computing”; “WoCo6 Programming Environments for High-Level Scientific Problem Solving”; and “WoCo8: Software Architectures for Scientific Computing Applications”.

The Working Conference is sponsored by the International Federation for Information Processing (IFIP) and organized by IFIP WG 2.5: Working Group on Numerical Software and the Center for Advanced Computing Research at the California Institute of Technology in cooperation with the Society for Industrial and Applied Mathematics (SIAM).

The organizational structure for the working conference is the following:

Chair

• James C. T. Pool, Retired, formerly Executive Director, Center for Advanced Computing Research, California Institute of Technology

Deputy Chair

• Brian Ford, Mathematics Faculty, Oxford University and Founding Director, Numerical Algorithms Group

Executive Committee

• Conference Chair
• Deputy Conference Chair
• Program Committee Chair
• WG2.5 Chair: Ronald F. Boisvert, Mathematical and Computational Sciences Division, National Institute of Standards and Technology

Program Committee

• Chair: William D. Gropp, Mathematics and Computer Science Division, Argonne National Laboratory
• Dennis Gannon, Department of Computer Science, Indiana University
• Jennifer Schopf, Mathematics and Computer Science Division, Argonne National Laboratory and UK National e-Science Centre, University of Edinburgh
• Masaaki Shimasaki, Department of Electrical Engineering, Kyoto University
• Michael ThunÈ, Department of Information Technology, Uppsala University
• Anne Trefethen; Interdisciplinary e-Research Centre; Oxford University
• Conference Chair
• Conference Deputy Chair
• Proceedings Co-Editors

Proceedings Co-Editors

• Conference Chair
• Patrick W Gaffney, Bergen Software Services International A/S, Bergen, Norway

Representative to the Society for Industrial and Applied Mathematics (SIAM)

• Conference Chair

Sponsored by

Organized by

In cooperation with

Supported by